bamec66557
commited on
Upload folder using huggingface_hub
Browse files- README.md +37 -226
- mergekit_config.yml +33 -103
- model-00001-of-00005.safetensors +1 -1
- model-00002-of-00005.safetensors +1 -1
README.md
CHANGED
@@ -1,119 +1,12 @@
|
|
1 |
---
|
2 |
-
|
|
|
|
|
3 |
library_name: transformers
|
4 |
tags:
|
5 |
- mergekit
|
6 |
- merge
|
7 |
-
|
8 |
-
base_model:
|
9 |
-
- bamec66557/VICIOUS_MESH-12B-BETA
|
10 |
-
- bamec66557/VICIOUS_MESH-12B-OMEGA
|
11 |
-
model-index:
|
12 |
-
- name: Mistral-Nemo-VICIOUS_MESH-12B-2407
|
13 |
-
results:
|
14 |
-
- task:
|
15 |
-
type: text-generation
|
16 |
-
name: Text Generation
|
17 |
-
dataset:
|
18 |
-
name: IFEval (0-Shot)
|
19 |
-
type: HuggingFaceH4/ifeval
|
20 |
-
args:
|
21 |
-
num_few_shot: 0
|
22 |
-
metrics:
|
23 |
-
- type: inst_level_strict_acc and prompt_level_strict_acc
|
24 |
-
value: 67.21
|
25 |
-
name: strict accuracy
|
26 |
-
source:
|
27 |
-
url: >-
|
28 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
29 |
-
name: Open LLM Leaderboard
|
30 |
-
- task:
|
31 |
-
type: text-generation
|
32 |
-
name: Text Generation
|
33 |
-
dataset:
|
34 |
-
name: BBH (3-Shot)
|
35 |
-
type: BBH
|
36 |
-
args:
|
37 |
-
num_few_shot: 3
|
38 |
-
metrics:
|
39 |
-
- type: acc_norm
|
40 |
-
value: 31.36
|
41 |
-
name: normalized accuracy
|
42 |
-
source:
|
43 |
-
url: >-
|
44 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
45 |
-
name: Open LLM Leaderboard
|
46 |
-
- task:
|
47 |
-
type: text-generation
|
48 |
-
name: Text Generation
|
49 |
-
dataset:
|
50 |
-
name: MATH Lvl 5 (4-Shot)
|
51 |
-
type: hendrycks/competition_math
|
52 |
-
args:
|
53 |
-
num_few_shot: 4
|
54 |
-
metrics:
|
55 |
-
- type: exact_match
|
56 |
-
value: 12.08
|
57 |
-
name: exact match
|
58 |
-
source:
|
59 |
-
url: >-
|
60 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
61 |
-
name: Open LLM Leaderboard
|
62 |
-
- task:
|
63 |
-
type: text-generation
|
64 |
-
name: Text Generation
|
65 |
-
dataset:
|
66 |
-
name: GPQA (0-shot)
|
67 |
-
type: Idavidrein/gpqa
|
68 |
-
args:
|
69 |
-
num_few_shot: 0
|
70 |
-
metrics:
|
71 |
-
- type: acc_norm
|
72 |
-
value: 8.84
|
73 |
-
name: acc_norm
|
74 |
-
source:
|
75 |
-
url: >-
|
76 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
77 |
-
name: Open LLM Leaderboard
|
78 |
-
- task:
|
79 |
-
type: text-generation
|
80 |
-
name: Text Generation
|
81 |
-
dataset:
|
82 |
-
name: MuSR (0-shot)
|
83 |
-
type: TAUR-Lab/MuSR
|
84 |
-
args:
|
85 |
-
num_few_shot: 0
|
86 |
-
metrics:
|
87 |
-
- type: acc_norm
|
88 |
-
value: 14.34
|
89 |
-
name: acc_norm
|
90 |
-
source:
|
91 |
-
url: >-
|
92 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
93 |
-
name: Open LLM Leaderboard
|
94 |
-
- task:
|
95 |
-
type: text-generation
|
96 |
-
name: Text Generation
|
97 |
-
dataset:
|
98 |
-
name: MMLU-PRO (5-shot)
|
99 |
-
type: TIGER-Lab/MMLU-Pro
|
100 |
-
config: main
|
101 |
-
split: test
|
102 |
-
args:
|
103 |
-
num_few_shot: 5
|
104 |
-
metrics:
|
105 |
-
- type: acc
|
106 |
-
value: 29.76
|
107 |
-
name: accuracy
|
108 |
-
source:
|
109 |
-
url: >-
|
110 |
-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bamec66557/Mistral-Nemo-VICIOUS_MESH-12B-2407
|
111 |
-
name: Open LLM Leaderboard
|
112 |
-
datasets:
|
113 |
-
- open-llm-leaderboard/bamec66557__Mistral-Nemo-VICIOUS_MESH-12B-2407-details
|
114 |
-
- open-llm-leaderboard/bamec66557__VICIOUS_MESH-12B-BETA-details
|
115 |
-
- open-llm-leaderboard/bamec66557__VICIOUS_MESH-12B-OMEGA-details
|
116 |
-
- open-llm-leaderboard/bamec66557__VICIOUS_MESH-12B-ALPHA-details
|
117 |
---
|
118 |
# merge
|
119 |
|
@@ -140,175 +33,93 @@ dtype: bfloat16
|
|
140 |
merge_method: slerp
|
141 |
tokenizer_source: base
|
142 |
|
143 |
-
# Slices Configuration
|
144 |
slices:
|
145 |
- sources:
|
146 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
147 |
-
layer_range: [0,
|
148 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
149 |
-
layer_range: [0, 5]
|
150 |
-
parameters:
|
151 |
-
t:
|
152 |
-
- name: self_attn
|
153 |
-
value: [0.85, 0.88, 0.91, 0.94, 0.97]
|
154 |
-
- name: mlp
|
155 |
-
value: [0.9, 0.92, 0.95, 0.98, 1.0]
|
156 |
-
- name: layer_norm
|
157 |
-
value: [0.75, 0.78, 0.81, 0.84, 0.87]
|
158 |
-
- name: embed_tokens
|
159 |
-
value: [1.0]
|
160 |
-
|
161 |
-
- sources:
|
162 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
163 |
-
layer_range: [5, 10]
|
164 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
165 |
-
layer_range: [5, 10]
|
166 |
-
parameters:
|
167 |
-
t:
|
168 |
-
- name: self_attn
|
169 |
-
value: [0.8, 0.83, 0.86, 0.89, 0.92]
|
170 |
-
- name: mlp
|
171 |
-
value: [0.88, 0.91, 0.94, 0.97, 1.0]
|
172 |
-
- name: layer_norm
|
173 |
-
value: [0.7, 0.73, 0.76, 0.79, 0.82]
|
174 |
-
- name: embed_tokens
|
175 |
-
value: [1.0]
|
176 |
-
|
177 |
-
- sources:
|
178 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
179 |
-
layer_range: [10, 15]
|
180 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
181 |
-
layer_range: [10, 15]
|
182 |
-
parameters:
|
183 |
-
t:
|
184 |
-
- name: self_attn
|
185 |
-
value: [0.75, 0.78, 0.81, 0.84, 0.87]
|
186 |
-
- name: mlp
|
187 |
-
value: [0.85, 0.88, 0.91, 0.94, 0.97]
|
188 |
-
- name: layer_norm
|
189 |
-
value: [0.65, 0.68, 0.71, 0.74, 0.77]
|
190 |
-
- name: embed_tokens
|
191 |
-
value: [1.0]
|
192 |
-
|
193 |
-
- sources:
|
194 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
195 |
-
layer_range: [15, 20]
|
196 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
197 |
-
layer_range: [
|
198 |
parameters:
|
199 |
t:
|
200 |
- name: self_attn
|
201 |
-
value: [0.
|
202 |
- name: mlp
|
203 |
-
value: [0
|
204 |
- name: layer_norm
|
205 |
-
value: [0.
|
206 |
-
- name: embed_tokens
|
207 |
-
value: [1.0]
|
208 |
|
209 |
- sources:
|
210 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
211 |
-
layer_range: [
|
212 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
213 |
-
layer_range: [
|
214 |
parameters:
|
215 |
t:
|
216 |
- name: self_attn
|
217 |
-
value: [0.
|
218 |
- name: mlp
|
219 |
-
value: [
|
220 |
- name: layer_norm
|
221 |
-
value: [0
|
222 |
-
- name: embed_tokens
|
223 |
-
value: [1.0]
|
224 |
|
225 |
- sources:
|
226 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
227 |
-
layer_range: [
|
228 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
229 |
-
layer_range: [
|
230 |
parameters:
|
231 |
t:
|
232 |
- name: self_attn
|
233 |
-
value: [0.
|
234 |
- name: mlp
|
235 |
-
value: [0.
|
236 |
- name: layer_norm
|
237 |
-
value: [0.
|
238 |
-
- name: embed_tokens
|
239 |
-
value: [1.0]
|
240 |
|
241 |
- sources:
|
242 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
243 |
-
layer_range: [30,
|
244 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
245 |
-
layer_range: [30,
|
246 |
parameters:
|
247 |
t:
|
248 |
- name: self_attn
|
249 |
-
value: [0.
|
250 |
- name: mlp
|
251 |
-
value: [0.
|
252 |
- name: layer_norm
|
253 |
-
value: [0.
|
254 |
-
- name: embed_tokens
|
255 |
-
value: [1.0]
|
256 |
-
|
257 |
-
- sources:
|
258 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
259 |
-
layer_range: [35, 40]
|
260 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
261 |
-
layer_range: [35, 40]
|
262 |
-
parameters:
|
263 |
-
t:
|
264 |
-
- name: self_attn
|
265 |
-
value: [0.6, 0.63, 0.66, 0.69, 0.72]
|
266 |
-
- name: mlp
|
267 |
-
value: [0.65, 0.68, 0.71, 0.74, 0.77]
|
268 |
-
- name: layer_norm
|
269 |
-
value: [0.4, 0.43, 0.46, 0.49, 0.52]
|
270 |
-
- name: embed_tokens
|
271 |
-
value: [1.0]
|
272 |
|
273 |
# Regularization
|
274 |
regularization:
|
|
|
|
|
275 |
- method: weight_clipping
|
276 |
-
clip_range: [-0.
|
277 |
- method: random_noise
|
278 |
-
scale: 0.
|
279 |
- method: attention_dropout
|
280 |
-
scale: 0.
|
281 |
-
- method: gradient_clipping
|
282 |
-
clip_norm: 1.0
|
283 |
|
284 |
# Postprocessing
|
285 |
postprocessing:
|
|
|
|
|
286 |
- operation: non_linear_scaling
|
287 |
parameters:
|
288 |
function: tanh
|
289 |
- operation: sharpening
|
290 |
-
intensity: 0.
|
291 |
- operation: gaussian_smoothing
|
292 |
-
sigma: 1.
|
293 |
- operation: normalize
|
294 |
- operation: dynamic_scaling
|
295 |
-
scale_range: [0.
|
296 |
- operation: smoothing
|
297 |
parameters:
|
298 |
adaptive: true
|
299 |
-
range: [0.
|
300 |
-
kernel_size:
|
301 |
|
302 |
```
|
303 |
-
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard)
|
304 |
-
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/bamec66557__Mistral-Nemo-VICIOUS_MESH-12B-2407-details)
|
305 |
-
|
306 |
-
| Metric |Value|
|
307 |
-
|-------------------|----:|
|
308 |
-
|Avg. |27.26|
|
309 |
-
|IFEval (0-Shot) |67.21|
|
310 |
-
|BBH (3-Shot) |31.36|
|
311 |
-
|MATH Lvl 5 (4-Shot)|12.08|
|
312 |
-
|GPQA (0-shot) | 8.84|
|
313 |
-
|MuSR (0-shot) |14.34|
|
314 |
-
|MMLU-PRO (5-shot) |29.76|
|
|
|
1 |
---
|
2 |
+
base_model:
|
3 |
+
- bamec66557/VICIOUS_MESH-12B-BETA
|
4 |
+
- bamec66557/VICIOUS_MESH-12B-OMEGA
|
5 |
library_name: transformers
|
6 |
tags:
|
7 |
- mergekit
|
8 |
- merge
|
9 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
---
|
11 |
# merge
|
12 |
|
|
|
33 |
merge_method: slerp
|
34 |
tokenizer_source: base
|
35 |
|
36 |
+
# Slices Configuration
|
37 |
slices:
|
38 |
- sources:
|
39 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
40 |
+
layer_range: [0, 10]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
42 |
+
layer_range: [0, 10]
|
43 |
parameters:
|
44 |
t:
|
45 |
- name: self_attn
|
46 |
+
value: [0.5, 0.55, 0.6, 0.65, 0.7]
|
47 |
- name: mlp
|
48 |
+
value: [1.0, 1.05, 1.1, 1.15, 1.2]
|
49 |
- name: layer_norm
|
50 |
+
value: [0.9, 0.95, 1.0, 1.05, 1.1]
|
|
|
|
|
51 |
|
52 |
- sources:
|
53 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
54 |
+
layer_range: [10, 20]
|
55 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
56 |
+
layer_range: [10, 20]
|
57 |
parameters:
|
58 |
t:
|
59 |
- name: self_attn
|
60 |
+
value: [0.4, 0.45, 0.5, 0.55, 0.6]
|
61 |
- name: mlp
|
62 |
+
value: [1.1, 1.15, 1.2, 1.25, 1.3]
|
63 |
- name: layer_norm
|
64 |
+
value: [1.0, 1.05, 1.1, 1.15, 1.2]
|
|
|
|
|
65 |
|
66 |
- sources:
|
67 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
68 |
+
layer_range: [20, 30]
|
69 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
70 |
+
layer_range: [20, 30]
|
71 |
parameters:
|
72 |
t:
|
73 |
- name: self_attn
|
74 |
+
value: [0.6, 0.65, 0.7, 0.75, 0.8]
|
75 |
- name: mlp
|
76 |
+
value: [0.9, 0.95, 1.0, 1.05, 1.1]
|
77 |
- name: layer_norm
|
78 |
+
value: [0.85, 0.9, 0.95, 1.0, 1.05]
|
|
|
|
|
79 |
|
80 |
- sources:
|
81 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
82 |
+
layer_range: [30, 40]
|
83 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
84 |
+
layer_range: [30, 40]
|
85 |
parameters:
|
86 |
t:
|
87 |
- name: self_attn
|
88 |
+
value: [0.7, 0.75, 0.8, 0.85, 0.9]
|
89 |
- name: mlp
|
90 |
+
value: [0.8, 0.85, 0.9, 0.95, 1.0]
|
91 |
- name: layer_norm
|
92 |
+
value: [0.8, 0.85, 0.9, 0.95, 1.0]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
|
94 |
# Regularization
|
95 |
regularization:
|
96 |
+
- method: gradient_penalty
|
97 |
+
scale: 0.05 # Increased influence for gradient control
|
98 |
- method: weight_clipping
|
99 |
+
clip_range: [-0.2, 0.2] # Broader clipping range for flexibility
|
100 |
- method: random_noise
|
101 |
+
scale: 0.01 # Stronger noise injection
|
102 |
- method: attention_dropout
|
103 |
+
scale: 0.1 # Higher dropout to reduce attention fixation
|
|
|
|
|
104 |
|
105 |
# Postprocessing
|
106 |
postprocessing:
|
107 |
+
- operation: entropy_regularization
|
108 |
+
scale: 0.05 # Stronger encouragement for diverse outputs
|
109 |
- operation: non_linear_scaling
|
110 |
parameters:
|
111 |
function: tanh
|
112 |
- operation: sharpening
|
113 |
+
intensity: 0.5 # Enhanced sharpening for precise outputs
|
114 |
- operation: gaussian_smoothing
|
115 |
+
sigma: 1.5 # Increased smoothing for stable outputs
|
116 |
- operation: normalize
|
117 |
- operation: dynamic_scaling
|
118 |
+
scale_range: [0.8, 1.2] # Expanded dynamic range for scaling
|
119 |
- operation: smoothing
|
120 |
parameters:
|
121 |
adaptive: true
|
122 |
+
range: [0.85, 1.15] # Wider adaptive smoothing range
|
123 |
+
kernel_size: 5
|
124 |
|
125 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mergekit_config.yml
CHANGED
@@ -3,161 +3,91 @@ dtype: bfloat16
|
|
3 |
merge_method: slerp
|
4 |
tokenizer_source: base
|
5 |
|
6 |
-
# Slices Configuration
|
7 |
slices:
|
8 |
- sources:
|
9 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
10 |
-
layer_range: [0,
|
11 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
12 |
-
layer_range: [0,
|
13 |
parameters:
|
14 |
t:
|
15 |
- name: self_attn
|
16 |
-
value: [0.
|
17 |
- name: mlp
|
18 |
-
value: [0
|
19 |
- name: layer_norm
|
20 |
-
value: [0.
|
21 |
-
- name: embed_tokens
|
22 |
-
value: [1.0]
|
23 |
|
24 |
- sources:
|
25 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
26 |
-
layer_range: [
|
27 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
28 |
-
layer_range: [
|
29 |
parameters:
|
30 |
t:
|
31 |
- name: self_attn
|
32 |
-
value: [0.
|
33 |
- name: mlp
|
34 |
-
value: [
|
35 |
- name: layer_norm
|
36 |
-
value: [0
|
37 |
-
- name: embed_tokens
|
38 |
-
value: [1.0]
|
39 |
|
40 |
- sources:
|
41 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
42 |
-
layer_range: [
|
43 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
44 |
-
layer_range: [
|
45 |
parameters:
|
46 |
t:
|
47 |
- name: self_attn
|
48 |
-
value: [0.
|
49 |
- name: mlp
|
50 |
-
value: [0.
|
51 |
- name: layer_norm
|
52 |
-
value: [0.
|
53 |
-
- name: embed_tokens
|
54 |
-
value: [1.0]
|
55 |
|
56 |
- sources:
|
57 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
58 |
-
layer_range: [
|
59 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
60 |
-
layer_range: [
|
61 |
parameters:
|
62 |
t:
|
63 |
- name: self_attn
|
64 |
-
value: [0.
|
65 |
- name: mlp
|
66 |
-
value: [0.8, 0.
|
67 |
- name: layer_norm
|
68 |
-
value: [0.
|
69 |
-
- name: embed_tokens
|
70 |
-
value: [1.0]
|
71 |
-
|
72 |
-
- sources:
|
73 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
74 |
-
layer_range: [20, 25]
|
75 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
76 |
-
layer_range: [20, 25]
|
77 |
-
parameters:
|
78 |
-
t:
|
79 |
-
- name: self_attn
|
80 |
-
value: [0.7, 0.73, 0.76, 0.79, 0.82]
|
81 |
-
- name: mlp
|
82 |
-
value: [0.75, 0.78, 0.81, 0.84, 0.87]
|
83 |
-
- name: layer_norm
|
84 |
-
value: [0.55, 0.58, 0.61, 0.64, 0.67]
|
85 |
-
- name: embed_tokens
|
86 |
-
value: [1.0]
|
87 |
-
|
88 |
-
- sources:
|
89 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
90 |
-
layer_range: [25, 30]
|
91 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
92 |
-
layer_range: [25, 30]
|
93 |
-
parameters:
|
94 |
-
t:
|
95 |
-
- name: self_attn
|
96 |
-
value: [0.68, 0.71, 0.74, 0.77, 0.8]
|
97 |
-
- name: mlp
|
98 |
-
value: [0.7, 0.73, 0.76, 0.79, 0.82]
|
99 |
-
- name: layer_norm
|
100 |
-
value: [0.5, 0.53, 0.56, 0.59, 0.62]
|
101 |
-
- name: embed_tokens
|
102 |
-
value: [1.0]
|
103 |
-
|
104 |
-
- sources:
|
105 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
106 |
-
layer_range: [30, 35]
|
107 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
108 |
-
layer_range: [30, 35]
|
109 |
-
parameters:
|
110 |
-
t:
|
111 |
-
- name: self_attn
|
112 |
-
value: [0.65, 0.68, 0.71, 0.74, 0.77]
|
113 |
-
- name: mlp
|
114 |
-
value: [0.68, 0.71, 0.74, 0.77, 0.8]
|
115 |
-
- name: layer_norm
|
116 |
-
value: [0.45, 0.48, 0.51, 0.54, 0.57]
|
117 |
-
- name: embed_tokens
|
118 |
-
value: [1.0]
|
119 |
-
|
120 |
-
- sources:
|
121 |
-
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
122 |
-
layer_range: [35, 40]
|
123 |
-
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
124 |
-
layer_range: [35, 40]
|
125 |
-
parameters:
|
126 |
-
t:
|
127 |
-
- name: self_attn
|
128 |
-
value: [0.6, 0.63, 0.66, 0.69, 0.72]
|
129 |
-
- name: mlp
|
130 |
-
value: [0.65, 0.68, 0.71, 0.74, 0.77]
|
131 |
-
- name: layer_norm
|
132 |
-
value: [0.4, 0.43, 0.46, 0.49, 0.52]
|
133 |
-
- name: embed_tokens
|
134 |
-
value: [1.0]
|
135 |
|
136 |
# Regularization
|
137 |
regularization:
|
|
|
|
|
138 |
- method: weight_clipping
|
139 |
-
clip_range: [-0.
|
140 |
- method: random_noise
|
141 |
-
scale: 0.
|
142 |
- method: attention_dropout
|
143 |
-
scale: 0.
|
144 |
-
- method: gradient_clipping
|
145 |
-
clip_norm: 1.0
|
146 |
|
147 |
# Postprocessing
|
148 |
postprocessing:
|
|
|
|
|
149 |
- operation: non_linear_scaling
|
150 |
parameters:
|
151 |
function: tanh
|
152 |
- operation: sharpening
|
153 |
-
intensity: 0.
|
154 |
- operation: gaussian_smoothing
|
155 |
-
sigma: 1.
|
156 |
- operation: normalize
|
157 |
- operation: dynamic_scaling
|
158 |
-
scale_range: [0.
|
159 |
- operation: smoothing
|
160 |
parameters:
|
161 |
adaptive: true
|
162 |
-
range: [0.
|
163 |
-
kernel_size:
|
|
|
3 |
merge_method: slerp
|
4 |
tokenizer_source: base
|
5 |
|
6 |
+
# Slices Configuration
|
7 |
slices:
|
8 |
- sources:
|
9 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
10 |
+
layer_range: [0, 10]
|
11 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
12 |
+
layer_range: [0, 10]
|
13 |
parameters:
|
14 |
t:
|
15 |
- name: self_attn
|
16 |
+
value: [0.5, 0.55, 0.6, 0.65, 0.7]
|
17 |
- name: mlp
|
18 |
+
value: [1.0, 1.05, 1.1, 1.15, 1.2]
|
19 |
- name: layer_norm
|
20 |
+
value: [0.9, 0.95, 1.0, 1.05, 1.1]
|
|
|
|
|
21 |
|
22 |
- sources:
|
23 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
24 |
+
layer_range: [10, 20]
|
25 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
26 |
+
layer_range: [10, 20]
|
27 |
parameters:
|
28 |
t:
|
29 |
- name: self_attn
|
30 |
+
value: [0.4, 0.45, 0.5, 0.55, 0.6]
|
31 |
- name: mlp
|
32 |
+
value: [1.1, 1.15, 1.2, 1.25, 1.3]
|
33 |
- name: layer_norm
|
34 |
+
value: [1.0, 1.05, 1.1, 1.15, 1.2]
|
|
|
|
|
35 |
|
36 |
- sources:
|
37 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
38 |
+
layer_range: [20, 30]
|
39 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
40 |
+
layer_range: [20, 30]
|
41 |
parameters:
|
42 |
t:
|
43 |
- name: self_attn
|
44 |
+
value: [0.6, 0.65, 0.7, 0.75, 0.8]
|
45 |
- name: mlp
|
46 |
+
value: [0.9, 0.95, 1.0, 1.05, 1.1]
|
47 |
- name: layer_norm
|
48 |
+
value: [0.85, 0.9, 0.95, 1.0, 1.05]
|
|
|
|
|
49 |
|
50 |
- sources:
|
51 |
- model: bamec66557/VICIOUS_MESH-12B-OMEGA
|
52 |
+
layer_range: [30, 40]
|
53 |
- model: bamec66557/VICIOUS_MESH-12B-BETA
|
54 |
+
layer_range: [30, 40]
|
55 |
parameters:
|
56 |
t:
|
57 |
- name: self_attn
|
58 |
+
value: [0.7, 0.75, 0.8, 0.85, 0.9]
|
59 |
- name: mlp
|
60 |
+
value: [0.8, 0.85, 0.9, 0.95, 1.0]
|
61 |
- name: layer_norm
|
62 |
+
value: [0.8, 0.85, 0.9, 0.95, 1.0]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
# Regularization
|
65 |
regularization:
|
66 |
+
- method: gradient_penalty
|
67 |
+
scale: 0.05 # Increased influence for gradient control
|
68 |
- method: weight_clipping
|
69 |
+
clip_range: [-0.2, 0.2] # Broader clipping range for flexibility
|
70 |
- method: random_noise
|
71 |
+
scale: 0.01 # Stronger noise injection
|
72 |
- method: attention_dropout
|
73 |
+
scale: 0.1 # Higher dropout to reduce attention fixation
|
|
|
|
|
74 |
|
75 |
# Postprocessing
|
76 |
postprocessing:
|
77 |
+
- operation: entropy_regularization
|
78 |
+
scale: 0.05 # Stronger encouragement for diverse outputs
|
79 |
- operation: non_linear_scaling
|
80 |
parameters:
|
81 |
function: tanh
|
82 |
- operation: sharpening
|
83 |
+
intensity: 0.5 # Enhanced sharpening for precise outputs
|
84 |
- operation: gaussian_smoothing
|
85 |
+
sigma: 1.5 # Increased smoothing for stable outputs
|
86 |
- operation: normalize
|
87 |
- operation: dynamic_scaling
|
88 |
+
scale_range: [0.8, 1.2] # Expanded dynamic range for scaling
|
89 |
- operation: smoothing
|
90 |
parameters:
|
91 |
adaptive: true
|
92 |
+
range: [0.85, 1.15] # Wider adaptive smoothing range
|
93 |
+
kernel_size: 5
|
model-00001-of-00005.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4865489336
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:606929c8ca1b67717338cc8a043d291796649691a67833de1c547f17f105b2ea
|
3 |
size 4865489336
|
model-00002-of-00005.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4907529456
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5d7ed0e2c571a185bafd363f79ddb85d760c1955008863eabef4fff22c42aa03
|
3 |
size 4907529456
|